Add RuntimeHelpers.IsReferenceOrContainsReferences<T>()

[jkotas]

Rename JitHelpers.ContainsReferences() to RuntimeHelpers.IsReferenceOrContainsReferences() and make it public.

Work towards https://github.com/dotnet/corefx/issues/14047

[jkotas]

cc @benaadams

I will fix this up if your change gets in first.

[benaadams]

is asm emitted for this for reference types (in current incarnation)? from #9540 (comment)

G_M58440_IG01:
       56                   push     rsi
       4883EC30             sub      rsp, 48
       48894C2428           mov      qword ptr [rsp+28H], rcx    ; putting something stack 40
       488BF1               mov      rsi, rcx                    ; putting it in rsi

G_M58440_IG02:
       488B0E               mov      rcx, qword ptr [rsi]        ;** ?? --- redundant load?        
       448B4618             mov      r8d, dword ptr [rsi+24]     ; int size = _size;
       33C9                 xor      ecx, ecx                    ;** 0 ----^ clears earlier move
       894E18               mov      dword ptr [rsi+24], ecx     ; _size = 0;
       FF461C               inc      dword ptr [rsi+28]          ; _version++;
       4585C0               test     r8d, r8d                    ; if (size > 0)
       7E15                 jle      SHORT G_M58440_IG04         ; ---> jmp  size == 0
       488B4E08             mov      rcx, gword ptr [rsi+8]      ; _items
       33D2                 xor      edx, edx                    ; 0
       488D0500000000       lea      rax, [(reloc)]              ; Array.Clear;

G_M58440_IG03:
       4883C430             add      rsp, 48
       5E                   pop      rsi
       48FFE0               rex.jmp  rax                         ; Array.Clear(_items, 0, size);

G_M58440_IG04:
       4883C430             add      rsp, 48
       5E                   pop      rsi
       C3                   ret      

There seems to be extra unnecessary register stuff going on; which could also be entirely unrelated. Also I'm not entirely fluent in asm or ABI so I might be wrong.

Was trying to work out if the startup part could be reduced any; but not sure how it relates to the code

[jkotas]

The redundant load is unused left over temp - opened #9547 on it. The unnecessary frame maybe side-effect of the unused left over temp.